Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print August 26, 2002, 10.1148/radiol.2251011090
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2251011090v1
225/1/151    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Arbab, A. S.
Right arrow Articles by Kumagai, H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Arbab, A. S.
Right arrow Articles by Kumagai, H.
(Radiology 2002;225:151-158.)
© RSNA, 2002

Gastrointestinal Imaging

Ferumoxides-enhanced Double-Echo T2-weighted MR Imaging in Differentiating Metastases from Nonsolid Benign Lesions of the Liver1

Ali S. Arbab, MBBS, PhD, Tomoaki Ichikawa, MD, Hironobu Sou, MD, Tsutomu Araki, MD, Hiroto Nakajima, MD, Keiichi Ishigame, MD, Takeharu Yoshikawa, MD and Hiroshi Kumagai, CRT

1 From the Department of Radiology, Yamanashi Medical University, Yamanashi, Japan. From the 2001 RSNA scientific assembly. Received June 21, 2001; revision requested August 15; final revision received April 1, 2002; accepted April 10. Address correspondence to A.S.A., LDRR/CC, National Institute of Health, Bldg 10, Rm B1N256, Bethesda, MD 20892 (e-mail: saali@cc.nih.gov).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 STATISTICAL CONSULTANT...
 REFERENCES
 
PURPOSE: To investigate whether ferumoxides-enhanced double-echo T2-weighted magnetic resonance (MR) imaging alone can allow differentiation of metastases from benign lesions in the noncirrhotic liver.

MATERIALS AND METHODS: At retrospective review of files and images, 60 lesions (22 metastases, 20 hemangiomas, and 18 cysts) were identified in 42 patients. All fast spin-echo T2-weighted MR images obtained before and after administration of ferumoxides with short (80–90 msec) and long (180–250 msec) echo times (TEs) were acquired with a 1.5-T system. Differences in lesion-to-liver signal intensity ratio between images obtained with long and short TEs were calculated. Data from all 60 lesions were entered into a receiver operating characteristic analysis. Three independent readers scored their observations of each lesion with a confidence level of 1–5. The diagnostic accuracy of each analysis method was determined by calculating the area under each reader-specific receiver operating characteristic curve. Interobserver agreement was calculated with the use of chance-corrected {kappa} statistics. Relative sensitivity, specificity, and accuracy of characterizing benign lesions with each method were calculated.

RESULTS: Markedly low signal intensity and lesion-to-liver ratio on ferumoxides-enhanced images were observed with hemangioma. The difference of lesion-to-liver ratio between long and short TEs on ferumoxides-enhanced images was significantly different from that of unenhanced images and that of metastases or cysts. Interobserver agreement was good to excellent. Ferumoxides-enhanced images (with short and long TEs) showed significantly higher diagnostic accuracy than that of unenhanced images (with short or short and long TEs). Ferumoxides-enhanced images showed similar sensitivity, specificity, and accuracy when all images were reviewed together.

CONCLUSION: Ferumoxides-enhanced T2-weighted MR images appear useful in differentiating metastases from benign (nonsolid) lesions in the liver.

© RSNA, 2002

Index terms: Liver, cysts, 761.3121 • Liver, hemangioma, 761.3194 • Liver neoplasms, metastases, 761.33 • Magnetic resonance (MR), contrast enhancement, 761.12143


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
 DISCUSSION
 STATISTICAL CONSULTANT...
 REFERENCES
 
Ferumoxides, one of the superparamagnetic iron oxide (SPIO) agents, has already shown its potential to enhance hepatocellular carcinoma and other focal lesions of the liver (1–5). There are reports (4,6–9) that show higher detectability of hepatic lesions with use of ferumoxides or SPIO, similar to that with use of computed tomography (CT) during arterial portography. It should also be emphasized, however, that ferumoxides presents major problems in characterizing hepatic lesions, especially those with low signal intensity (SI) on T1-weighted images. After administration of ferumoxides, the SI of the liver becomes lower with all sequences, which in turn causes higher SI of the lesions with less or no uptake of ferumoxides when compared with that of the surrounding liver parenchyma.

It is well known that the hemodynamic evaluation of hepatic lesions with contrastmaterial–enhanced dynamic magnetic resonance (MR) imaging often increases qualitative lesion characterization (10–14). The usefulness of bolus injection of one of the SPIOs (SHU-555A) in combination with echo-planar sequences in assessing lesion vascularity has been confirmed in previous studies (15,16). Currently, however, the only commercially available SPIO is not recommended as a bolus agent because of possible dose-dependent acute hypotensive reactions that are probably due to transient microvascular embolization by ferumoxides particles (1). As a result, it is difficult to characterize hepatic lesions with the use of ferumoxides on the basis of differences in vascularity of the lesions.

To resolve this limitation of lesion characterization with ferumoxides, unenhanced MR images obtained with various sequences are always read with ferumoxides-enhanced images for interpretation (17). T2-weighted MR images are of great value in characterizing different types of hepatic lesions, but these results are often limited by variability and overlap among different types of hepatic lesions, particularly between hemangioma and metastasis (18–20). Moreover, hemangioma and metastasis may often coexist in a patient with a noncirrhotic liver. It is sometimes difficult to differentiate these two types of lesions on unenhanced MR images alone, because both hemangioma and metastasis show a similar pattern of SI. In addition, if the lesions are not demonstrated on unenhanced MR images, characterization with ferumoxides-enhanced MR images alone may be difficult. Even the combination of both unenhanced and ferumoxides-enhanced images has shown variable accuracy in the differentiation of hepatic lesions (17,21). Moreover, acquisition of both unenhanced and SPIO-enhanced images would prolong investigation time. Therefore, it would be advantageous if only ferumoxides-enhanced MR images obtained with different imaging sequences could allow differentiation of various hepatic lesions.

Ferumoxides-enhanced images show strong T2 shortening and susceptibility effects in the liver because of ferumoxides accumulated in Kupffer cells and in the bloodstream. However, the susceptibility effect of ferumoxides will depend on the length of the echo time (TE) of T2-weighted images. A report (17) has shown improvement of characterization of hepatic lesions on SPIO-enhanced MR images obtained with double-echo T2-weighted imaging. In the same way, we hypothesized that T2-weighted imaging with longer TEs than those used in the previously reported study may improve differentiation of metastases from hemangiomas because of increased susceptibility effect for the presence of iron particles within hemangiomas. Thus, the purpose of this study was to investigate whether ferumoxides-enhanced double-echo T2-weighted MR images alone allow accurate differentiation of metastases from benign lesions (hemangiomas and cysts) in the noncirrhotic liver.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 STATISTICAL CONSULTANT...
 REFERENCES
 
Patient Population and Selection Criteria
A retrospective review of surgical, pathologic, and radiologic records at our institute identified 38 patients with hepatic hemangiomas and 36 patients with hepatic metastases from February 1998 to October 2000. The study design was approved by the hospital ethics committee, and proper consent was obtained from each patient. Review of records and images was also approved. Eighteen patients with hepatic hemangiomas were excluded from the study because they were not followed up more than 1 year after the initial diagnosis. Fourteen patients with hepatic metastases were excluded because there were no pathologic findings for at least one representative lesion.

Forty-two patients were included in the study, none of which had underlying cirrhosis. In these 42 patients, 25 lesions suggestive of cysts were also identified. Ten of the 42 patients had multiple lesions (multiple hemangiomas in four patients, multiple metastases in six patients, and multiple cysts in seven patients). However, only one lesion representative of a hemangioma, metastasis, or cyst was selected in each patient. These lesions included 20 hepatic hemangiomas in 20 patients, 22 hepatic metastases in 22 patients, and 18 cysts in 18 patients (10 with hemangiomas, eight with metastases). All 60 lesions included in the assessment were more than 1 cm in diameter. Because of the presence of double lesions in 18 patients, correlation was assessed by using paired data (lesion-to-liver SI ratios of metastasis vs cyst and hemangioma vs cyst) to see the dependency on subsequent statistical analysis. There was negligible correlation among these lesions.

Hemangiomas
For the 20 patients with hemangiomas (nine men, 11 women), patient age at diagnosis ranged from 28 to 81 years (mean, 51 years). Diagnosis of all cases of hemangioma was based on previously established characteristic findings at ultrasonography (US) and CT. The lesions that were hyperechoic at US, had low attenuation at nonenhanced CT with centripetal and gradual enhancement after contrast material administration, and did not change in size or shape for at least 1 year at contrast-enhanced dynamic CT were considered hemangiomas. From the first diagnosis of hemangioma, multiple follow-up sessions of US, CT, and MR imaging were repeated regularly in all patients for more than 1 year. On MR images, the mean maximum diameter of the selected hemangioma was 3.5 cm (range, 1.2–6.5 cm).

Metastases
For 22 cases of metastases (12 men, 10 women), patient age at diagnosis ranged from 35 to 75 years (mean, 53 years). The diagnosis of hepatic metastases was confirmed histopathologically in all patients by examining at least one representative lesion by means of either hepatic surgery (n = 8) or needle biopsy (n = 14). Of 22 cases of metastases, 10 were from colon carcinoma, eight were from gastric cancer, and four were from pancreatic carcinoma. In all patients with metastases, the appearance of new lesions and/or growth in the representative lesions was present at subsequent noninvasive follow-up examination at 2–8 months. In all metastases, the hypovascular nature of the lesions was suggested at dynamic CT and/or digital subtraction angiography. On MR images, the mean maximum diameter of selected metastases was 3.8 cm (range, 1.1–6.0 cm).

Cysts
For 18 cysts (10 men, eight women), patient age at diagnosis ranged from 35 to 71 years (mean, 54 years). The diagnosis of cysts was confirmed by means of US, CT, and clinical follow-up. The lesions that were hypoechoic at US, had low attenuation without enhancement at contrast-enhanced CT, and did not change in size or shape for at least 1 year were considered cysts. On MR images, the mean maximum diameter of the selected cysts was 3 cm (range, 1.5–4.0 cm).

There were no significant differences in patient age, lesion size (Mann-Whitney test, P > .1), and male to female ratio ({chi}2 test, P > .1) among hemangiomas, metastases, and cysts. All patients underwent both T1- and T2-weighted unenhanced and ferumoxides-enhanced MR imaging, but only T2-weighted images obtained before and after ferumoxides administration were assessed for this study.

MR Imaging
All studies were performed with a 1.5-T superconducting system (Signa Horizon; GE Medical Systems, Milwaukee, Wis). In all patients, both pre- and postcontrast T2-weighted images were obtained with a fast spin-echo technique (echo train length, 11), with different repetition times and TEs. Images with both short (TE, 80–90 msec; bandwidth, 31.2 kHz) and long (TE, 180–250 msec; bandwidth, 10.4 kHz) TEs were acquired. Repetition times were variable and depended on the respiratory gating. All sections were 7 mm thick, with a 3-mm intersection gap and a 256 x 128 matrix. Field of view was variable and ranged from 32 to 40 cm. After precontrast images were acquired, 0.05 mg/kg of ferumoxides (Feridex; Tanabe Pharmaceutical, Tokyo, Japan) was administered as a slow drip over 30 minutes. Postcontrast images were acquired within 60 minutes (45–60 minutes) after the start of administration of ferumoxides.

Region of Interest Setting and Collection of Data
A radiologist (H.S.) who was actively involved in the acquisition of images but was not involved in image analysis (or receiver operating characteristic [ROC] analysis) used a workstation (Advantage Window; GE Medical Systems) to create a region of interest over the lesions and normal hepatic parenchyma on pre- and postcontrast MR images (T2-weighted images with short and long TEs), and average SI was noted. The region of interest for each lesion was carefully placed within the confines of the entire lesion. As a rule, for heterogeneous lesions, the regions of interest were placed to include the entire lesion, regardless of differing SI within the lesions. The lesion-to-liver SI ratio (average SI of lesion [T] divided by average SI of liver [L] = T/L) was then calculated. The following formula was applied to calculate the SI change in each lesion between short and long TEs for both pre- and postcontrast images: T/L(long TE) - T/L(short TE).

The acquired data were expressed as mean ± standard error of the mean and presented graphically. The significant difference was assessed by means of analysis of variance, followed by the post hoc (Fisher PLSD) test.

Image Interpretation and Analysis
Both pre- and postcontrast T2-weighted images were analyzed. Two radiologists (T.I., H.S.), serving as coordinators, initially reviewed all images with knowledge of clinical-pathologic findings. On the basis of the clinical-pathologic findings, they attempted to determine the diagnosis of all lesions in all patients. Then they selected sections of pre- and postcontrast images containing the representative lesions to be presented for analysis. Each section contained only one type of representative lesion. If initial images were too poor in quality to interpret, new images were reprinted by the coordinators.

Three independent experienced abdominal radiologists (K.I., H.N., T.Y.) were asked to interpret the sections containing the lesions. The three readers were blinded to patient history and had not participated in consensus readings of the images. The sections obtained in a patient were not given to the readers all at the same time. The study coordinators presented the MR images separately and randomly to each reader (randomized by patient and sequence of images).

The following images were assessed by the readers: (a) precontrast T2-weighted images with short TEs, (b) precontrast T2-weighted images with both short and long TEs, (c) both pre- and postcontrast T2-weighted images with short TEs, (d) postcontrast T2-weighted images with both short and long TEs, and (e) all pre- and postcontrast images with short and long TEs. The points analyzed were (a) predominant SI of the lesions (such as high attenuation, isoattenuation, or low attenuation) compared with that of the surrounding liver tissue, (b) homogeneity of the lesion (homogeneous or heterogeneous), and (c) change of SI (markedly decreased, slightly decreased, or no change) between images obtained with short and long TEs after ferumoxides administration.

ROC Analysis
Data from all 60 lesions were entered into an ROC analysis. The same three independent readers were also asked to score their observations of each lesion with a confidence level of 1 to 5 (1 = definitely benign, 2 = probably benign, 3 = equivocal, 4 = probably malignant, and 5 = definitely malignant) during the assessment of the images. The lesions were considered benign when they showed homogeneously high SI (similar to that of water) on precontrast images, with either markedly decreased SI with the use of long TEs or no change at all after ferumoxides administration. The lesions were considered malignant when they showed mild to moderately high SI, heterogeneity, and slightly decreased SI or no change at all on long TE images obtained after ferumoxides administration.

Statistical Analysis
For each analysis method (precontrast short TE, precontrast short and long TE, pre- and postcontrast short TE, postcontrast short and long TE, and all combined pre- and postcontrast methods), a binormal ROC curve was fitted to each reader’s confidence scoring data by using a maximum likelihood estimation (22). The diagnostic accuracy of each method was determined by calculating the area under each reader-specific ROC curve (Az) (23). Composite ROC curves used to represent the performance of the three readers as a group were calculated by averaging the binormal parameter values of individual curves for each method. To find the significant differences among the methods and readers (multireader-multimodality ROC analysis), the jackknife method of Dorfman et al (24) was applied. Any lesion with a confidence level of 1 or 2 was considered benign. Relative sensitivity, specificity, and accuracy in characterizing benign lesions with each method were calculated for each reader. The values of relative sensitivity, specificity, and accuracy were transformed to normality by means of arcsin transformation to make the proportions satisfy analysis of variance assumptions. Repeated measures—analysis of variance, followed by a post hoc (Fisher PLSD) test—were applied to find any significant differences among the methods for the relative sensitivity, specificity, and accuracy.

All data were expressed as mean ± SD, unless otherwise indicated. The calculated SI change for each type of lesion between short and long TEs on both pre- and postcontrast images was analyzed by means of analysis of variance, followed by the post hoc (Fisher PLSD) test. A P value of less than .05 was considered to indicate a statistically significant difference.

For qualitative image interpretation, interobserver agreement for each kind of MR image was assessed for establishing the reliability of image interpretation. The degree of interobserver agreement between each combination of two readers was calculated with chance-corrected {kappa} statistics. Generally, a {kappa} value greater than 0.75 is considered to indicate excellent agreement beyond chance; 0.40–0.75 indicates fair to good agreement; and less than 0.40 indicates poor agreement.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 STATISTICAL CONSULTANT...
 REFERENCES
 
None of the selected image sections were discarded for poor image quality or for any other reason. During the analysis, most of the benign lesions (cysts and hemangiomas) showed homogeneously high SI on precontrast short and long TE images when compared with that of the surrounding normal liver tissue. Postcontrast long TE images showed marked decreased SI for hemangiomas (Fig 1). On the other hand, cysts did not show any decreased SI on postcontrast long TE images (Fig 2). Most of the metastatic lesions showed slightly to moderately high SI and inhomogeneity on precontrast short and long TE images when compared with that of the surrounding liver tissue, with equivocal to mild decreased SI on postcontrast long TE images (Fig 2).



View larger version (154K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Representative case of hemangioma. A, Precontrast short TE MR image shows typical homogeneously high SI in the hemangioma, with slightly decreased SI on B, the precontrast long TE MR image. C, Postcontrast short TE MR image shows comparatively more decreased SI than that of the surrounding liver. D, Postcontrast long TE MR image shows marked decrease in SI, although surrounding liver becomes almost black.

 


View larger version (163K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Representative cases of metastasis and cysts in one patient. A, Precontrast short TE, B, precontrast long TE, C, postcontrast short TE, and D, postcontrast long TE MR images. The metastasis (thick arrow) shows inhomogeneously high SI, with slightly decreased SI on long TE images for both pre- and postcontrast long TE sequences. The cyst (thin arrow) shows no definite change of homogeneous SI on pre- and postcontrast short or long TE images.

 
Figure 3 shows the difference of lesion-to-liver SI ratio between long and short TEs on pre- and postcontrast images. Because of accumulation of ferumoxides in Kupffer cells, the SI of hepatic tissue became very low, causing increased lesion-to-liver ratio for cysts, though no significant difference was found in the lesion-to-liver ratio difference between pre- and postcontrast images. Because of the presence of vascular structures in the hemangiomas, however, pooling of ferumoxides in the blood caused markedly decreased SI in hemangiomas, thereby significantly decreasing lesion-to-liver ratio on postcontrast images. The difference of lesion-to-liver ratio with long and short TEs on postcontrast images became inverse and significantly different (P < .001) from that of precontrast images. Though metastasis also showed a decrease in difference of lesion-to-liver ratio on postcontrast images, the difference was not significant (P = .107) when compared with that of precontrast images. The differences in lesion-to-liver ratio between long and short TEs within each type of lesion (cysts, hemangiomas, and metastases) were significant (P < .001) on pre- and postcontrast images.



View larger version (25K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Graph shows the differences (mean ± standard error of the mean) of lesion-to-liver SI ratios between short and long TE images for pre- and postcontrast sequences. Note the significant reversal of lesion-to-liver SI ratio differences in hemangioma. # = Significant differences (P < .001) between pre- and postcontrast sequences. * = significant differences (P < .001) among hemangiomas, cysts, and metastases.

 
Table 1 shows the interobserver agreement determined by means of {kappa} statistics in analyzing the images. Excellent agreement was observed among the readers when all pre- and postcontrast images were reviewed together. Better agreement was observed among the readers for pre- and postcontrast images with short and long TEs. Figure 4 shows the composite ROC curves and Table 2 shows the Az indices of individual readers for each method of analysis. No significant difference was found among readers by means of jackknife analysis. All readers improved their performance in characterizing lesions when long and short TE images were analyzed together. The lowest accuracy of lesion characterization was observed with precontrast short TE images. Although analysis of all pre- and postcontrast images together showed the highest accuracy in characterizing lesions, no significant differences were observed when compared with that of precontrast images (short and long TEs) and postcontrast images (short and long TEs). The Az index for the three readers was 0.88 ± 0.01 for precontrast short TE images, 0.94 ± 0.02 for combined precontrast short and long TE images, 0.89 ± 0.01 for combined pre- and postcontrast short TE images, 0.94 ± 0.02 for combined postcontrast short and long TE images, and 0.94 ± 0.02 for all pre- and postcontrast images.


View this table:
[in this window]
[in a new window]

  		TABLE 1. {kappa} Values among the MR Image Readers

 


View larger version (30K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4. Graph shows composite ROC curves in the detection of benign lesions of the liver. Az values are presented in Table 2. Both precontrast short TE images and pre- and postcontrast short TE images show significant differences with other images. TPF = true-positive fraction, FPF = false-positive fraction, Pre-S = precontrast short TE images, Pre-S+L = precontrast short and long TE images, Pre+Post-S = pre and postcontrast short TE images, Post-S+L = postcontrast short and long TE images, Pre+Post-S+L = pre- and postcontrast short and long TE images.

 

View this table:
[in this window]
[in a new window]

  		TABLE 2. Az Indices of Individual Readers for Each Method of Image Analysis

 
Figure 5 shows the cumulative relative sensitivity, specificity, and accuracy (data represent arcsin transformation of the proportions) in characterizing benign lesions for different methods of analysis. The highest sensitivity and accuracy were found when all the images (pre- and postcontrast) were considered together. The highest specificity was found when precontrast short and long TE images were considered together, but significant differences were not found among the analysis methods. When postcontrast T2-weighted long and short TE images were considered, significantly higher relative sensitivity and accuracy were observed than those with precontrast long and short TE images or with combined pre- and postcontrast short TE images. The lowest relative sensitivity and accuracy were observed when precontrast T2-weighted images with short and long TEs were reviewed together.



View larger version (49K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5. Graph shows the relative sensitivity, specificity, and accuracy in characterizing benign lesions (hemangiomas and cysts) of the liver. The values of the proportions are transformed to normality by means of arcsin transformation (see statistical commentary). The data are expressed as mean ± 95% CI. Pre-s = precontrast T2-weighted images with short TEs, Pre-SL = precontrast T2-weighted images with short and long TEs, Post-SL = postcontrast T2-weighted images with short and long TEs, Pre-post-S = pre- and postcontrast T2-weighted images with short TEs, Pre-Post-SL = pre- and postcontrast T2-weighted images with short and long TEs, a = significant differences compared with pre- and postcontrast T2-weighted images with short and long TEs, b = significant differences compared with postcontrast T2-weighted images with short and long TEs.

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
STATISTICAL CONSULTANT...
 REFERENCES
 
Ferumoxides and other SPIOs have already shown their potential to enhance all kinds of focal lesions of the liver (1–9). To our knowledge, investigators in previous studies (1–9) used various sequences of unenhanced and ferumoxides- and/or SPIO-enhanced images for the detection and characterization of hepatic lesions. A few investigators (18,20,25,26) pointed out that a long TE for T2-weighted MR images could allow differentiation of solid from nonsolid lesions of the liver, but they did not use short and long TE SPIO-enhanced T2-weighted images. T2-weighted MR images are of great value in characterizing different types of hepatic lesions, but these results are often limited by variability and overlap among different types of hepatic lesions, particularly between hemangioma and metastasis (20,26,27). Hemangioma and metastasis may often coexist in a patient with a noncirrhotic liver. The use of both pre- and postcontrast single-echo T2-weighted images showed variable accuracy in allowing differentiation of hepatic lesions (17,21). Moreover, obtaining both unenhanced and SPIO-enhanced images is time consuming. In the present study, we tried to confirm whether postcontrast T2-weighted images (with long and short TEs) alone could allow differentiation of benign (nonsolid) from malignant lesions in the noncirrhotic liver.

In our study, the quality of the images obtained with fast spin-echo sequences was satisfactory in all cases, with use of both long and short TEs. Similar to findings in previous investigations, most of the metastatic lesions showed mild to moderate SI loss on precontrast long TE images, whereas cysts and hemangiomas did not show any significant SI loss on long TE images. This might be due to inherent T2 relaxation properties of the lesions. Investigators in previous reports (18,20,26) pointed out that the T2 relaxation of cysts and hemangiomas was slower than that of metastases and solid tumors of the liver.

In the present study, we did not measure the T2 relaxivity of the lesions; instead, we quantitatively measured the lesion-to-liver SI ratio differences between short and long TEs on images obtained before and after ferumoxides administration. In our investigation, only patients with noncirrhotic livers were enrolled, so it was assumed that the decrease of hepatic SI after ferumoxides administration would not be significantly different from patient to patient. Significant reversal of lesion-to-liver SI ratio difference for hemangioma on postcontrast images compared with that on precontrast images indicates a significant decrease in SI on long TE images obtained after ferumoxides administration. This might be due to the susceptibility effect of the accumulated ferumoxides in hemangiomas (pooling in the blood), causing marked decrease in SI on long TE images. Although not significant, the decrease in lesion-to-liver SI ratio difference of metastasis on postcontrast images indicates the relative vascularity of metastasis. On the other hand, cysts showed slightly nonsignificant increased lesion-to-liver SI ratio differences on postcontrast images, which indicate an absence of any effect of ferumoxides on the SI of the cysts. This finding also indicates the absence of vascular structures within the cysts and is useful in differentiating them from those of hemangiomas and metastases. Differences of lesion-to-liver SI ratio between short and long TE postcontrast images might be useful in differentiating hemangiomas, metastases, and cysts of the liver.

Excellent agreement among readers, higher Az values at ROC analysis, and the highest sensitivity and accuracy for combined pre- and postcontrast images clearly indicates the usefulness of ferumoxides in characterizing lesions of the liver. However, acquisition of both pre- and postcontrast double-echo T2-weighted images is time consuming, and the protocol would not increase the patient throughput. Although pre- and postcontrast T2-weighted images (both types obtained with short and long TEs) showed similar higher Az values at ROC analysis, the sensitivity and accuracy in characterizing lesions were significantly lower with precontrast images (with short and long TEs). This indicates the advantage of using postcontrast images (with short and long TEs) over precontrast images. On the other hand, there were no significant differences observed for sensitivity, specificity, or accuracy in characterizing lesions of the liver between the image review methods of combined postcontrast short and long TE T2-weighted images and all combined pre- and postcontrast T2-weighted images. Although there are reports that show improved detection and characterization of lesions of the liver by using two different TEs for T2-weighted images, no report has been found that uses SPIO and double-echo T2-weighted images in the analysis of hepatic lesions (18,20). Grangier et al (28) pointed out the significant decrease in SI of hemangiomas on SPIO-enhanced double-echo T2-weighted images, but their study was limited to SI changes, and they did not try to detect or characterize the hepatic lesions.

Our results, obtained with precontrast short and long TE T2-weighted images, support the results of Fenlon et al (20), but significantly higher sensitivity and accuracy with postcontrast short and long TE T2-weighted images in characterizing lesions of the liver indicate the usefulness of postcontrast images. Moreover, nonsignificant differences (although larger sample size might show significant differences) for sensitivity, specificity, accuracy, and Az values between the postcontrast short and long TE T2-weighted images and all pre- and postcontrast short and long TE T2-weighted images indicate that postcontrast short and long TE T2-weighted images alone might be sufficient to characterize hepatic lesions in patients with noncirrhotic livers. Acquiring only postcontrast short and long TE T2-weighted images will shorten the total investigation time and increase the throughput of patients in any MR imaging division.

The purpose of this study was to characterize lesions (hemangiomas, cysts, and metastases) of the liver on T2-weighted images by using two different TEs before and after ferumoxides administration and to examine the diagnostic qualities of postcontrast T2-weighted images (with short and long TEs) alone over either precontrast or combined pre- and postcontrast T2-weighted images. T1-weighted images were not considered during characterization of the lesions in the present study, but in clinical practice, a T1-weighted image might be needed in case of hemangioma with thrombosis or fibrosis. Moreover, no patients had hepatocellular carcinoma in our study, because most of the patients with hepatocellular carcinoma usually have underlying cirrhosis, which would increase nonspecific overestimation of hepatic lesions after ferumoxides administration due to the presence of regenerative or dysplastic nodules. To overcome this false-positive lesion detection, T1-weighted images or precontrast T2-weighted images would be needed.

In conclusion, ferumoxides-enhanced T2-weighted images (obtained with short and long TEs) appear useful in the differentiation of metastases from benign (nonsolid) lesions in the noncirrhotic liver.


   STATISTICAL CONSULTANT COMMENTARY
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 STATISTICAL CONSULTANT...
REFERENCES
 
In this article, the authors used sensitivity, specificity, and accuracy as data in an analysis of variance. Such analytic methods generally require that the data follow a normal distribution and that the variance in each group is the same. Unfortunately, sensitivity, specificity, and accuracy are proportions, and the variance of a proportion is highly dependent on the value of the proportion, particularly if that proportion is close to 0 or 1.0. One solution in such cases is to "transform" the data prior to analysis in such a way that the transformed data meet the assumptions for the analysis. When dealing with proportions, the inverse trigonometric function "arcsin" transformation is usually recommended. If p is our proportion, then the transformed data would be x, where x is chosen such that sin(x) = {surd}p (Statistical methods. 6th ed. Ames, Iowa: Iowa State University Press, 1974). The original values of sensitivity, specificity, or accuracy were between 0 and 1. The transformed values, x, lie between 0° and 90° if an angular measurement is used or between 0 and 1.57 ({pi}/2) if radians are used, as the authors did in this case. This transformation, while used primarily for variance stabilization, also makes the data appear more normally (ie, Gaussian) distributed. Ultimately, the statistical test will be more likely to indicate differences if they exist.


   FOOTNOTES
 
Abbreviations: Az = area under the ROC curve, ROC = receiver operating characteristic, SI = signal intensity, SPIO = superparamagnetic iron oxide, TE = echo time

Author contributions: Guarantors of integrity of entire study, A.S.A., T.I., T.A.; study concepts and design, A.S.A., T.I., H.S.; literature research, A.S.A., T.I., H.S.; clinical studies, T.I., H.S., H.K., A.S.A.; data acquisition, A.S.A., T.I., H.K.; data analysis/interpretation, H.S., T.I., H.N., K.I., T.Y.; statistical analysis, A.S.A., T.I., K.I.; manuscript preparation, A.S.A., T.I.; manuscript definition of intellectual content, A.S.A., T.I., T.A.; manuscript editing and revision/review, T.I., T.A.; manuscript final version approval, all authors.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 STATISTICAL CONSULTANT...
 REFERENCES
 

  1. Stark DD, Weissleder R, Elizondo G, et al. Superparamagnetic iron oxide: clinical application as a contrast agent for MR imaging of the liver. Radiology 1988; 168:297-301.[Abstract/Free Full Text]
  2. Weissleder R, Hahn PF, Stark DD, et al. Superparamagnetic iron oxide: enhanced detection of focal splenic tumors with MR imaging. Radiology 1988; 169:399-403.[Abstract/Free Full Text]
  3. Oudkerk M, van den Heuvel AG, Wielopolski PA, Schmitz PI, Borel Rinkes IH, Wiggers T. Hepatic lesions: detection with ferumoxides-enhanced T1-weighted MR imaging. Radiology 1997; 203:449-456.[Abstract/Free Full Text]
  4. Seneterre E, Taourel P, Bouvier Y, et al. Detection of hepatic metastases: ferumoxides-enhanced MR imaging versus unenhanced MR imaging and CT during arterial portography. Radiology 1996; 200:785-792.[Abstract/Free Full Text]
  5. Ros PR, Freeny PC, Harms SE, et al. Hepatic MR imaging with ferumoxides: a multicenter clinical trial of the safety and efficacy in the detection of focal hepatic lesions. Radiology 1995; 196:481-488.[Abstract/Free Full Text]
  6. Lencioni R, Donati F, Cioni D, Paolicchi A, Cicorelli A, Bartolozzi C. Detection of colorectal liver metastases: prospective comparison of pre- and postcontrast magnetic resonance imaging at 1.5 T, dual-phase spiral CT, and spiral CT during arterial portography. MAGMA 1998; 7:76-87.
  7. Strotzer M, Gmeinwieser J, Schmidt J, et al. Diagnosis of liver metastases from colorectal adenocarcinoma: comparison of spiral-CTAP combined with intravenous contrast-enhanced spiral-CT and SPIO-enhanced MR combined with plain MR imaging. Acta Radiol 1997; 38:986-992.[Medline]
  8. Semelka RC, Schlund JF, Molina PL, et al. Malignant liver lesions: comparison of spiral CT arterial portography and MR imaging for diagnostic accuracy, cost, and effect on patient management. J Magn Reson Imaging 1996; 6:39-43.[Medline]
  9. Choi D, Kim S, Lim J, et al. Preoperative detection of hepatocellular carcinoma: ferumoxides-enhanced MR imaging versus combined helical CT during arterial portography and CT hepatic arteriography. AJR Am J Roentgenol 2001; 176:475-482.[Abstract/Free Full Text]
  10. Ohtomo K, Itai Y, Yoshikawa K, et al. Hepatic tumors: dynamic MR imaging. Radiology 1987; 163:27-31.[Abstract/Free Full Text]
  11. van Beers B, Demeure R, Pringot J, et al. Dynamic spin-echo imaging with Gd-DTPA: value in the differentiation of hepatic tumors. AJR Am J Roentgenol 1990; 154:515-519.[Abstract/Free Full Text]
  12. Mano I, Yoshida H, Nakabayashi K, Yashiro N, Iio M. Fast spin echo imaging with suspended respiration: gadolinium-enhanced MR imaging of liver tumors. J Comput Assist Tomogr 1987; 11:73-80.[Medline]
  13. Murakami T, Mitani T, Nakamura H, et al. Differentiation between hepatoma and hemangioma with inversion-recovery snapshot FLASH MRI and Gd-DTPA. J Comput Assist Tomogr 1992; 16:198-205.[Medline]
  14. Schmiedl U, Kolbel G, Hess CF, Klose U, Kurt B. Dynamic sequential MR imaging of focal liver lesions: initial experience in 22 patients at 1.5 T. J Comput Assist Tomogr 1990; 14:600-607.[Medline]
  15. Ichikawa T, Arbab AS, Araki T, et al. Perfusion MR imaging with a superparamagnetic iron oxide using T2-weighted and susceptibility-sensitive echoplanar sequences: evaluation of tumor vascularity in hepatocellular carcinoma. AJR Am J Roentgenol 1999; 173:207-213.[Abstract/Free Full Text]
  16. Arbab AS, Ichikawa T, Araki T, et al. Detection of hepatocellular carcinoma and its metastases with various pulse sequences using superparamagnetic iron oxide (SHU-555-A). Abdom Imaging 2000; 25:151-158.[CrossRef][Medline]
  17. Reimer P, Jahnke N, Feibich M, et al. Hepatic lesion detection and characterization: value of nonenhanced MR imaging, superparamagnetic iron oxide-enhanced MR imaging, and spiral CT-ROC analysis. Radiology 2000; 217:152-158.[Abstract/Free Full Text]
  18. McFarland EG, Mayo-Smith WW, Saini S, Hahn PF, Goldberg MA, Lee MJ. Hepatic hemangiomas and malignant tumors: improved differentiation with heavily T2-weighted conventional spin-echo MR imaging. Radiology 1994; 193:43-47.[Abstract/Free Full Text]
  19. Kanematsu M, Hoshi H, Murakami T, et al. Focal hepatic lesion detection: comparison of four T2-weighted MR imaging pulse sequences. Radiology 1998; 206:167-175.[Abstract/Free Full Text]
  20. Fenlon HM, Tello R, deCarvalho VL, Yucel EK. Signal characteristics of focal liver lesions on double echo T2-weighted conventional spin echo MRI: observer performance versus quantitative measurements of T2 relaxation times. J Comput Assist Tomogr 2000; 24:204-211.[CrossRef][Medline]
  21. Bluemke DA, Paulson EK, Choti MA, DeSena S, Clavien PA. Detection of hepatic lesions in candidates for surgery: comparison of ferumoxides-enhanced MR imaging and dual-phase helical CT. AJR Am J Roentgenol 2000; 175:1653-1658.[Abstract/Free Full Text]
  22. Metz CE. ROC methodology in radiologic imaging. Invest Radiol 1986; 21:720-733.[Medline]
  23. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982; 143:29-36.[Abstract/Free Full Text]
  24. Dorfman DD, Berbaum KS, Metz CE. Receiver operating characteristics rating analysis: generalization of the population of readers and patients with the jackknife method. Invest Radiol 1992; 27:723-731.[CrossRef][Medline]
  25. Ito K, Mitchell DG, Outwater EK, Szklaruk J, Sadek AG. Hepatic lesions: discrimination of nonsolid, benign lesions from solid, malignant lesions with heavily T2-weighted fast spin echo MR imaging. Radiology 1997; 204:729-737.[Abstract/Free Full Text]
  26. Ohtomo K, Itai Y, Furui S, Yashiro N, Yoshikawa K, Iio M. Hepatic tumors: differentiation by transverse relaxation time (T2) of magnetic resonance imaging. Radiology 1985; 155:421-423.[Abstract/Free Full Text]
  27. Ohtomo K, Itai Y, Yoshida H, Kokubo T, Yoshikawa K, Iio M. MR differentiation of hepatocellular carcinoma from cavernous hemangioma: complementary roles of FLASH and T2 values. AJR Am J Roentgenol 1989; 152:505-507.[Abstract/Free Full Text]
  28. Grangier C, Tournaire J, Mentha G, et al. Enhancement of hepatic hemangioma on T1-weighted MR SE images by superparamagnetic iron oxide particles. J Comp Assist Tomogr 1994; 18:888-896.[Medline]



This article has been cited by other articles:


Home page
 Am. J. Roentgenol.Home page
K. W. Kim, A. Y. Kim, T. K. Kim, S. H. Park, H. J. Kim, Y. K. Lee, M.-S. Park, H. K. Ha, P. N. Kim, J. C. Kim, et al.
Small (l.e. 2cm) Hepatic Lesions in Colorectal Cancer Patients: Detection and Characterization on Mangafodipir Trisodium-Enhanced MRI
Am. J. Roentgenol., May 1, 2004; 182(5): 1233 - 1240.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2251011090v1
225/1/151    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Arbab, A. S.
Right arrow Articles by Kumagai, H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Arbab, A. S.
Right arrow Articles by Kumagai, H.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE